Characteristics of Patients with Diabetes Initiating Sodium Glucose Co-transporter-2 Inhibitors (SGLT2i): Real-World Results from Three Administrative Databases in Japan

This study retrieved patient data from three distinct administrative databases in Japan: (1) a hospital-based administrative database (H-dataset) constructed from data on inpatients and outpatients from 287 Diagnosis Procedure Combination (DPC) hospitals [20]; (2) a pharmacy claims database (P-dataset) comprising data from over 800 pharmacies nation-wide, which provided coverage for approximately 2% of all outpatient prescriptions [21]; and (3) an insurance claims database (I-dataset) containing medical and prescription claims of 3.8 million employees and their dependents (as of study execution) who were mostly aged ≤ 65 years [22].

This study included patients from each database who were ≥ 18 years old as of 1 April 2014 and initiated therapy on an OAD between 1 April 2014 and 31 March 2017 (study period) (Fig. 1). Patients who had any of the following International Classification of Diseases tenth revision (ICD-10) diagnosis codes during the study period were excluded: E10.xx (type 1 diabetes mellitus), E12.xx (malnutrition-related diabetes mellitus), E13.xx (other specified diabetes mellitus) and O24.xx (diabetes mellitus in pregnancy). Patients with any SGLT2i prescription were identified, and the first medication date (= index date) was noted. Patients who initiated treatment with SGLT2i and any other OAD(s) or with ≥ 2 SGLT2i agents on the same date were excluded from the study. The remaining patients, consisting of those who added SGLT2i to their prior OAD therapy or those who started SGLT2i monotherapy, were included in the SGLT2i cohort.

Since the other OADs included in this study for comparison have been in the market for a long time, we tried to include new users during the study period to allow better comparison with the SGLT2i cohort. Other OAD cohorts in this study included patients receiving therapy with α-GI, BG, DPP-4i, glinides, SU, and TZD [see Electronic Supplementary Material (ESM) Table 1 for a list of medication codes]. We identified the first prescription date for each OAD class and flagged this date as a candidate index date. If the patients used the index OAD during the 6-month pre-index period or initiated therapy with the index OAD together with another class of OADs or fixed-dose combination drugs on the same date (co-initiation), then the candidate index OAD was excluded. Finally, the earliest candidate was selected as the index OAD, and the initial prescription date for the index OAD was identified as the index date. Patients who had < 6 months enrollment prior to the index date were excluded from the study cohorts (only applicable to the I-dataset). For better generalizability, patients who were hospitalized at the index date were excluded from the analysis (only applicable to the H- and I-datasets).

We evaluated patient characteristics and prescribing site characteristics for the SGLT2i and the other OAD cohorts. A window period of − 30 days was allowed for the collection of baseline clinical values [body mass index (BMI), glycated hemoglobin (HbA1c), estimated glomerular filtration rate (eGFR)]. If there were multiple values within this period, the closest one to the index date was chosen. Comorbidities were coded according to the Elixhauser Comorbidity Index (ECI) [23, 24] and scored as previously reported [25]. The prevalence of hypertension and hypercholesterolemia were assessed based on the ICD-10 diagnosis code (I10.x for hypertension and E78.x for hypercholesterolemia) and prescriptions for these conditions during the pre-index period (YJ codes starting with 214 for hypertension and 218 for hypercholesterolemia). Diabetes-related complications were evaluated using the Diabetes Complication Severity Index (DCSI) [26], which identified seven complications: CVD, nephropathy, retinopathy, cerebrovascular disease, neuropathy, peripheral vascular disease and metabolic disease (ketoacidosis). The DCSI was originally developed using the ICD ninth revision (ICD-9) codes and laboratory test results, but ICD-10 codes [27, 28] with adaptations for Japan-specific standardized diagnosis codes and serum creatinine results (where available) were used in this study (see ESM Table 2 for a list of diagnosis codes).

We analyzed the data in each dataset separately. Continuous data were presented as the mean and standard deviation (SD) or as the median and inter-quartile range (IQR), as appropriate, and categorical data were presented as proportion. Prescription site was summarized according to the number of beds (≥ 20 beds as hospital and < 20 beds as clinic). Changes over time in prescription site, prescribing specialty and age at index were aggregated quarterly by index date. All analyses were conducted for each dataset separately using SAS^® Studio Release 3.5 (SAS Institute Inc., Cary, NC, USA).

In total, 176,355 patients in the H-dataset, 98,361 patients in the P-dataset, and 37,786 patients in the I-datasets were included in the analysis (Fig. 1). Table 1 shows the characteristics of SGLT2i users and of users of other OADs. Overall, 38.8% of the patients in the P-dataset and 68.7% of those in the I-dataset received their index prescriptions from clinics. Age/gender distribution were similar between patients in the H-dataset (only included hospital patients) and those in the P-dataset (included both clinic and hospital patients). The H-dataset contained more prescriptions initiated by diabetologists than did the P-dataset. Compared with users of other OADs, SGLT2i users and TZD users tended to have more prescriptions initiated from the clinics. Over the study period, there was a decreasing trend in SGLT2i prescriptions initiated from the clinics (ESM Fig. 1). OAD prescriptions were most frequently initiated by generalists across all cohorts (Table 1). Notably, the number of SGLT2i prescriptions initiated in the cardiology department increased slightly over the study period (ESM Fig. 2).

There were no differences in gender distribution between SGLT2i users and users of other OADs in each dataset, but SGLT2i users were the youngest among the OAD cohorts (Table 1; Fig. 2). Over the study period, there was an increasing trend in mean age at index among SGLT2i users (Fig. 3), with an increase in the proportion of patients aged ≥ 65 years in all datasets (ESM Fig. 3).

The comorbidities and diabetes-related complications at index among SGLT2i users and users of other OADs in the H- and I-datasets are summarized in Table 2. The median ECI score was lower in both SGLT2i and BG users than in users of other OADs in the H-dataset but was comparable across all OAD users in the I-dataset. A higher prevalence of hypertension was noted in SGLT2i users in the I-dataset and a higher prevalence of hypercholesterolemia was noted in SGLT2i users in both the H- and I-datasets compared with users of other OADs. The median DCSI score was similar across all OAD cohorts in both the H- and I-datasets. There were no distinct differences in the prevalence of complications between SGLT2i users and users of other OADs in both of these datasets.

The body weight, BMI, HbA1c and eGFR levels at index in SGLT2i users and users of other OADs in the H- and I-datasets are shown in Table 3. SGLT2i users in the H-dataset had the highest mean body weight, BMI and HbA1c levels among the OAD cohorts (Table 3). The mean BMI level was also highest in SGLT2i users in the I-dataset. All OAD cohorts in both datasets had median eGFR levels of > 60 mL/min/1.73 m².